Sparse linear algebraic methods and MPI for spatio-temporal inference

Environmental systems are characterised by phenomena which evolve both in space and in time (cloud modelling, weather and climate, glacial flow, spread of invasive species etc.).

The aim of spatio-temporal modelling is to

• Assimilate all knowledge (observables etc.) pertaining to the process/es of interest in an objective way (although some of the beliefs we incorporate might be subjective).
• Infer parameters governing the spatio-temporal process in order to
  • Further our understanding of the process and
  • Predict in both space and time.

• Geo-statistical models:

\[ X_t(\svec) \sim GP(\mu_t(\svec),K(t,t',\svec,\svec')) \]

General-purpose model, used when the dynamic structure is largely unknown.

• Dynamic (mechanistically-motivated) models

\[ X_t(\svec) = f(X_{t-1}(\svec), X_{t-2}(\svec), \dots, \wvec_t(\svec); \thetab_m) \]

• Examples:
  • STAR models
  • Integro-difference equation models
  • PDE models
  • CMLs

• Used when we have some mechanistic insight of the underlying process.
• In the environmental sciences these are usually in the form of (stochastic) PDEs.
• They implicitly define very complicated spatio-temporal covariances.

\[ \frac{\partial X(s;t)}{\partial t} - \beta \frac{\partial^2X(s;t)}{\partial s^2} + a X(s;t) = e(s;t) \]

• Continuous-time ST processes are computationally tedious to deal with.

• Focus on models of the form

\[ X_t(\svec) = f(X_{t-1}(\svec)) + e_t(\svec) \]

where \( ~e_t(\svec)~ \) is the solution to

\[ (\kappa^2 - \Delta)^{\alpha/2}(\tau e(\svec)) = \mathcal{W(\svec)} \]

\[ (\kappa^2 - \Delta)^{\alpha/2}(\tau e(\svec)) = \mathcal{W(\svec)} \]

has as solution (Whittle, 1954)

\[ e(\svec) \sim \mathcal{GP}(0,K(\svec,\svec'; \alpha,\tau,\kappa)) \]

where \( k(\cdot,\cdot) \) is a Matérn covariance function. The wave spectrum of \( e(\svec) \) is given by

\[ R(\kvec) = (2\pi)^{-\alpha}(\kappa^2 + \| \kvec \|^2)^{-\alpha} \]

1. Non-stationarity can be handled by making the SPDE spatially varying. This results in a covariance function which is not a Matern field, but do we care?
   

2. We can model SPDEs on manifolds, not restricted to \( \mathbb{R}^2 \).
   

3. SPDEs can be approximated to arbitrary precision using an appropriate basis (Lindgren, 2011).

\[ X_{t+1}(\svec) = \mathcal{A}X_t(\svec) + e_t{\svec} \]

1. Project:

\[ \langle \phib,X_{t+1}(\svec) \rangle = \langle \phib,\mathcal{A}X_{t+1}(\svec) \rangle + \langle \phib,e_t(\svec)\rangle \]

1. Interpolate:

\[ \langle \phib,\psib^T \rangle\xvec_{t+1} = \langle \phib,\mathcal{A}\psib^T \rangle\xvec_t + \langle \phib,\psib^T\rangle\evec_t \]

For \( \alpha = 2 \) and \( \psib = \phib \) (Galerkin method):

\[ [ \langle\phib, \kappa^2\phib^T\rangle - \langle\phib, \Delta\phib^T\rangle)]\tau \evec = \langle \phib,\mathcal{W(\svec)} \rangle \]

• Useful when we have multiple interacting spatio-temporal processes

• Consider the bi-variate sub-system of random forcings: \( e_{1,t}(\svec), e_{2,t}(\svec) \). Define the multi-variate Gaussian process (MVGP) as

\[ (e_{1,t}(\svec), e_{2,t}(\svec)) = \bold{e}(\svec) \sim \mathcal{MVGP}(\bold{0},\bold{K}(\svec,\svec',t,t';\thetab_K)) \]

where \( \bold{K}(\svec,\svec',t,t';\thetab_K) \) is a \( 2 \times 2 \) cross-covariance matrix function (i.e.~ each element of \( \bold{K} \) is a covariance function (Finley, 2007):

\[ \bold{K}(\svec,\svec',t,t') = [Cov(e_{i,t}(\svec), e_{i,t'}(\svec'))]_{i,j \in \{1,2\}} \]

• Model with a system of SPDEs. Example of a three-variate system of random forcings:

\[ \begin{bmatrix} \mathcal{A}_1 & & \\ & & \mathcal{A}_2 & \mathcal{A}_{2,3} \\ & & \mathcal{A}_{3,2} & \mathcal{A}_3 \end{bmatrix} \begin{bmatrix} e_1(\svec) \\ e_2(\svec) \\ e_3(\svec) \end{bmatrix} = \begin{bmatrix} \mathcal{W}_1(\svec) \\ \mathcal{W}_2(\svec) \\ \mathcal{W}_3(\svec) \end{bmatrix} \]

• SPDEs are a useful way to modelling complex phenomena.

• Real power comes in using finite elements to decompose and represent them as GMRFs

• From now on we restrict ourselves to the linear, Gaussian case

\[ \begin{bmatrix} \xvec_{1,t} \\ \xvec_{2,t} \\ \vdots \end{bmatrix} = \begin{bmatrix} \Amat_{1} & \Amat_{1,2} & \dots \\ \Amat_{2,1}& \Amat_{2,2} & \dots \\ \vdots& \vdots & \ddots \\ \end{bmatrix} \begin{bmatrix} \xvec_{1,t-1} \\ \xvec_{2,t-1} \\ \vdots \end{bmatrix} + \bold{J}_t\bold{\beta} + \begin{bmatrix} \evec_{1,t} \\ \evec_{2,t} \\ \vdots \end{bmatrix} \]

• Kalman filter/ RTS methods do not lend themselves to sparse methods (more of that in another talk)

• Collapse all variates, and their spatial and temporal components into one big GMRF

• Sparsity is retained and one can make use of fill-in reduction permutations when computing the Cholesky factor (Knorr-Rue and Held, 2002)

\[ p(\xvec,\betab_I | \Thetab) = \left(\prod_{i=1}^T{p(\xvec_{t} | \xvec_{t-1},\betab,\Thetab)}\right)p(\xvec_0 | \betab, \Thetab)p(\betab_I) \]

• Initial distribution (Note, not the stationary one!):

\[ \xvec_0|\betab_I,\Thetab \sim \mathcal{N}(\bold{J}_0\betab,[\Qmat_w - \Amat^T\Qmat_w\Amat]^{-1}) \]

\[ \small \begin{array}{cc} \Qmat = \begin{bmatrix} \Qmat_1 & \Qmat_2 \\ \Qmat_2^T & \Qmat_3 \end{bmatrix} & \Qmat_2 = \begin{bmatrix} \Amat^T\Qmat_w\Jmat_{1} - \Qmat_0\Jmat_{0} \\ \Amat^T\Qmat_w\Jmat_{2} - \Qmat_w\Jmat_1 \\ \vdots\end{bmatrix}, \end{array} \]

\[ \small \Qmat_1 = \begin{bmatrix} \Amat^T\Qmat_w\Amat + \Qmat_0 & -\Amat^T\Qmat_w & ~~~ \\ -\Qmat_w\Amat & \Amat^T\Qmat_w\Amat + \Qmat_w& -\Amat^T\Qmat_w & ~~~ \\ & -\Qmat_w\Amat & \ddots & \ddots \\ & \ddots & \ddots & \ddots & \end{bmatrix} \]

\[ \small \Qmat_3 = \Jmat_0^T\Qmat_0\Jmat_0 + \sum_{i=1}^{T} (\Jmat_i^T\Qmat_w\Jmat_i) + \Qmat_\beta \]

• Conditioned on the parameters, this is the simplest model possible.
• \( \Qmat \) is large. Example: Antarctica, 4 ST fields, 7 time points, 500k observations, 500k vertices.
• Updated \( \Qmat^* \approx 3.5Gb \). \( \Qmat^{*-1} \approx 1Tb \)
• What if we want to find \( diag(\Amat \Sigma^* \Amat^T) \) where \( \Sigma^* = \Qmat^{*-1} \)?
• Impossible to carry out on a standard PC for large, \( \Amat \), even if \( chol(\Qmat^*) \) is known and fill-in reduced.

• A full backsolve for \( \Amat \Sigma^* \Amat^T \) is wasteful since \( \Amat \) is very sparse.
• We only need a few elements of \( \Sigma^* \). Which elements are these?

Lemma 1: Takahashi Equations (Erisman and Tinney, 1975) Let \( \Lmat = chol(\Qmat) \). If \( L_{ij} \ne 0 \), then \( \Sigma_{ij} \) can be computed using elements in

• \( \Lmat \) and
• \( \Sigma^*_{IJ}, I = \{ i'; i' > i\}, J = \{j'; j' > j\} \)

• A full backsolve for \( \Amat \Sigma^* \Amat^T \) is wasteful since \( \Amat \) is very sparse.
• We only need a few elements of \( \Sigma^* \). Which elements are these?

Theorem 1: Let \( \yvec = \Cmat\xvec + \vvec \) and \( \vvec \sim \mathcal{N}(0,\Tmat^{-1}) \). Then if

• \( \textrm{zeros}(\Amat) \subseteq \textrm{zeros}(\Cmat) \)
• \( \Amat \) is non-negative
• \( \Tmat \) is diagonal, then \( diag(\Amat \Sigma^* \Amat^T) \equiv diag(\Amat \Sigma^*_{00} \Amat^T) \)

• The updated precision matrix \( \Qmat^* = \Qmat + \Amat^T\Tmat^{-1}\Amat \) is sparse if both \( \Qmat \) and \( \Amat \) are sparse and \( \Tmat \) is diagonal.

Lemma 2: If \( \Amat \) is non-negative then \( [\Amat^T\Amat]_{jk} = 0 \Leftrightarrow diag(\Amat\Sigma^*\Amat^T) \) does not depend on \( \Sigma^*_{jk} \)

• Let \( zeros(\cdot) \) denote the symbolic pattern of a matrix.

\[ \begin{align} zeros(\Qmat^*) &= zeros(\Qmat + \Amat^T\Tmat^{-1}\Amat) \\ &\ge zeros(\Amat^T\Tmat^{-1}\Amat) = zeros(\Amat^T\Amat) \end{align} \]

1. In practice the Takahashi equations are always used with GMRFs in order to obtain the marginal variances. Theorem 2 provides a means to obtain uncertainty on linear combinations essentially for free.

2. I can provide predictive uncertainties at any point provided that each row of \( zeros(\Amat) \subseteq zeros(\Cmat) \).

• In all practical cases there are unknown parameters and a massive GMRF, sampling may be required
• MCMC with multiple chains is relatively straightforward
  • e.g. Evolutionary MCMC, Embarassingly parallel MCMC
• Extremely little work on parallelising a single chain
• Take advantage of local Markov property

\[ x_i \perp \xvec_{-\{i,ne(i)\}} | \xvec_{ne(i)} ~~~ \forall i \in \mathcal{V} \]

\[ E(x_i | \xvec_{-i}) = -\frac{1}{Q_{ii}}\sum_{j:j\sim i}Q_{ij}x_j; ~~~ prec(x_i | \xvec_{-i}) = Q_{ii} \]

• Single-site updating in spatio-temporal systems is a bad idea
• We can make use of the global Markov property

\[ \xvec_A \perp \xvec_B | \xvec_{C} \]

\( \forall A,B,C \notin \varnothing \) and \( C \) separating \( A,B \).

• In typical Bayesian networks, finding \( A, B \) and \( C \) can be difficult (seed and spawn idea, Gonzales, 2011).

• With a spatial network, this is remarkably easy using the Kernighan-Lin algorithm for graph partitioning (Kernighan and Lin, 1970).

• Multi-variate spatio-temporal statistics has a lot to offer in a large range of environmental problems which remain untapped.

• Computational properties of GMRFs allow us to consider major sources of uncertainty with influence on experimental design: e.g. in multivariate spatial problems, coverage is not as important as mixture diversity

• Paramter estimation with large-scale spatio-temporal problems is hard but usually possible (INLA, VB, MCMC), but with large-scale multi-variate spatio-temporal under-determined problems considerably harder. Parallel sampling offers a possible solution.

• Parallel sampling can also become intractable, what about parallel approximate Bayesian sparse methods? http://arxiv.org/abs/1305.4152